Magnetic particle imaging (MPI) is a young and developing volumetric imaging technique first proposed in 2005 that directly detects the magnetization from magnetic particles without depth limitations. MPI has important applications to medical imaging, e.g., heart and blood vessel imaging, cell tracking, and cancer detection. The basic principle of MPI involves exciting magnetic particles in a selected region (e.g., magnetic particle contrast agents injected into the blood stream or labeled into or on cells) and detecting their non-linear response. An inhomogeneous magnetic field having a field-free region and a strong field outside this region selects the particles that will be detected. Such a field may be established, for example, using two permanent magnets with opposing magnetic field orientations. Magnetic particles located outside the field-free region become magnetically saturated by the strong field. Non-saturated magnetic particles located inside the field-free region are excited by generating an oscillating magnetic field superimposed on the inhomogeneous field. The oscillating field may be generated, for example, using solenoids. The excited particles are detected by measuring their non-linear response to the oscillating field. Because tissue has a negligible nonlinear response, detecting the nonlinear harmonics from the magnetic particles provides a high contrast signal. The magnetic particles are normally composed of a non-linear ferromagnetic material, e.g., ferumoxide or super-paramagnetic iron oxide (SPIO). Volumetric imaging is performed by changing the relative position of the field-free region with respect to the distribution of the magnetic particles (e.g., by altering the inhomogeneous field to displace the field-free region and/or moving the object being imaged).
In MPI techniques proposed by Gleich and Weizenecker (US 2003/0085703, WO 2004/091395, WO 2008/099331), the magnetic field used to excite the magnetic particles oscillates at a frequency f0 in the radiofrequency range, and signals at a series of harmonics at 2f0, 3f0, 4f0, . . . , 20f0, are measured using a receiver coil. Because these harmonics span a large bandwidth, detecting them poses challenges for the receiver design. For example, with f0=25.5 kHz, the harmonics span a large 500 kHz bandwidth. Gleich et al. use an un-tuned receiver coil, which can not be optimally matched to the preamplifier, to detect the signals over this large bandwidth.
Gleich and Weizenecker teach a technique for creating an image from a set of N harmonics of the fundamental excitation frequency. However, there are several challenges with their technique. First, their method casts the image data over a very broad frequency range, precluding narrow-band detection and associated benefits. Also, their image reconstructed from the N harmonic images is not optimal. Moreover, conventional deconvolution algorithms can lead to pernicious noise amplification when applied outside of their original context of a single image blurred by a single point spread function. Thus, it would be an advance in the state of the art to provide a method for combining N images from separate harmonics to create a single high resolution image with minimal noise amplification.
In the method of Gleich and Weizenecker, each high resolution pixel is sampled, requiring Mx*My*Mz samples, where Mx is the number of voxels in x direction, My is the number of voxels in the y direction, and MZ is the number of voxels in the z direction. However, the magnetic particle contrast agent is likely to be substantially sparse, meaning many of these voxels are substantially void of contrast and, hence, signal. Consequently, the sampling method is inefficient and slow. It would thus be an advance in the state of the art to provide more efficient methods for MPI imaging.